One video in three languages: a timeline job per language

Render each language as its own Timeline 1.0 job: same video[] slots, a different Sume-hosted audio spine. The render is $0.10 per output minute with ffmpeg only, so three 60-second versions cost $0.30 plus the speech synthesis for each language.

What stays and what changes

The video slots, transitions and output size stay identical. Only audio.url and audio.duration_seconds change, because a translated voice is rarely the same length. Declared video[].start values are authoritative, so if a language runs longer you must lengthen the slots too.

Loop over languages

Every URL must already be on media.sume.com. The audio URLs below are placeholders for TTS results you already have; the slot timings are for a 24-second spine.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
VOICE = {
    "en": os.environ["VOICE_EN"],
    "es": os.environ["VOICE_ES"],
    "ko": os.environ["VOICE_KO"],
}
for lang, url in VOICE.items():
    body = {
        "audio": {"url": url, "duration_seconds": 24},
        "video": [{"source_url": os.environ["CLIP_URL"], "start": 0, "duration": 24}],
    }
    r = requests.post("https://api.sume.com/v1/timeline-1.0/render",
                      headers={**H, "Idempotency-Key": f"loc-render-{lang}-001"},
                      json=body, timeout=30)
    r.raise_for_status()
    print(lang, r.json()["status_url"])

Then captions

Burn per-language captions with cues in a separate video-captions job on each render, using a Hangul style for Korean. Lip movement still follows the source language; for on-camera speakers that is a separate lip-sync step.